In September 2026, Yakov Pachochich, chief scientist at OpenAI, published an article revealing that they were advancing their Recursive Self-Improvement project, which is regarded as the core of next-generation general artificial intelligence. The team used coding agent technology to significantly accelerate the research process. Data shows that AI spending per researcher increased significantly by the end of July, presumably due to internal employees gaining access to the GPT-6 Astra model at that time. Internal research at OpenAI indicates that this technology has had a significant impact on experimental speed, task complexity, and overall research efficiency, making 2026 a year of major breakthroughs in this field.
In September 2026, OpenAI launched the recursive self-improvement project and allowed employees to access the GPT-6 Astra model.
Coverage · reports per dayLANGUAGE SPLIT
Entity relations
Integrated timelineUNIFIED TIMELINE
2026-09-06
OpenAI disclosed progress in the recursive self-improvement project.
Chief scientist Jakub Pachoci published an article discussing the recursive self-improvement project, and the team used a coding agent to accelerate the research and development process.
The internal R&D coding agent of OpenAI has revolutionized artificial intelligence research. Early data shows that this technology has had a significant impact on the speed of experiments, the complexity of tasks, and the acceleration of research.
Jakub Pachocki, Chief Scientist at OpenAI, published a new article titled “Alien Thinking,” in which he discusses his Recursive Self-Improvement project, which is regarded as his new generation of general artificial intelligence. The OpenAI research team is using coded agents to accelerate development progress, and 2026 is expected to be a year of significant breakthroughs in this field. Data shows that per-person AI spending increased significantly by the end of July, presumably due to internal employees gaining access to the GPT-6 Astra model at that time.